Machine learning approaches for enhanced estimation of reference evapotranspiration (ETo): a comparative evaluation.

Area of Science:

• Hydrology
• Agricultural Science
• Machine Learning

Background:

• Accurate reference evapotranspiration (ETo) estimation is crucial for water resource management, especially in data-limited areas.
• Existing models often require extensive data or complex methodologies, hindering their use in resource-constrained environments.

Purpose of the Study:

• To evaluate static machine learning models (KNN, DT, RF) for daily ETo estimation using varied input scenarios.
• To identify the most effective model and key meteorological predictors for ETo estimation in data-scarce regions.

Main Methods:

• Application of K-Nearest Neighbors (KNN), Decision Tree (DT), and Random Forest (RF) models.
• Temporal dependency analysis to justify static model application.
• ETo estimation using full-feature and minimal single-variable input datasets.

Main Results:

• Random Forest (RF) demonstrated superior performance, achieving a root mean square error (RMSE) of 0.52 mm/day and R² of 0.96.
• Temperature and solar radiation were identified as the most significant predictors for ETo estimation.
• RF model showed robustness across different input data scenarios.

Conclusions:

• Static machine learning models, particularly RF, offer a practical and efficient approach for ETo estimation in data-scarce regions.
• The findings provide a reliable tool for enhancing water management and agricultural planning under limited data conditions.